[petsc-users] seg fault with superlu

Thu Dec 6 17:53:10 CST 2018

We just fixed a memory bug in 6.1.0 (which shows up in very rare
situation), not sure whether it is related to your problem.

But "-mat_superlu_dist_colperm NATURAL" will be very slow, usually generate
lots of fill-ins in the LU factors.

Sherry

On Thu, Dec 6, 2018 at 3:50 PM Marius Buerkle via petsc-users <
petsc-users at mcs.anl.gov> wrote:

> Thanks for the suggestion I will try this option. I used already
> "-mat_superlu_dist_colperm NATURAL" which seems to help but I did not test
> it thoroughly.
>
>
> *Betreff:* Re: [petsc-users] seg fault with superlu
>
> On Thu, Dec 6, 2018 at 2:55 AM Matthew Knepley via petsc-users <
> petsc-users at mcs.anl.gov> wrote:
>
>> Satish,
>>
>> This is exactly the same method (pzgssvx) that had a problem before. Do
>> you remember what happened?
>>
>
> Used to have a bug in the row permutation algorithm. The row permutation
> is not get populated correctly. It was fixed from PETSc-3.8.x.   If the
> current issue is the same as before, a simple "fix" is to use
>
> " -mat_superlu_dist_fact SamePattern_SameRowPerm".
>
> Fande,
>
>
>
>>
>>   Thanks,
>>
>>      Matt
>>
>> On Wed, Dec 5, 2018 at 8:38 PM Smith, Barry F. via petsc-users <
>> petsc-users at mcs.anl.gov> wrote:
>>
>>>
>>>    I would start with valgrind see if it finds anything. If not and if
>>> you can provide the code that reproduces the problem we can take a look at
>>> it.
>>>
>>>     Barry
>>>
>>> https://www.mcs.anl.gov/petsc/documentation/faq.html#valgrind
>>>
>>>
>>>
>>> > On Dec 5, 2018, at 7:16 PM, Marius Buerkle via petsc-users <
>>> petsc-users at mcs.anl.gov> wrote:
>>> >
>>> > Hi,
>>> >
>>> > I get the following error duing MatMatSolve. It works for smaller
>>> number of processes but if I  increade the number pf processes used it
>>> crashes.
>>> >
>>> > [25]PETSC ERROR:
>>> ------------------------------------------------------------------------
>>> > [25]PETSC ERROR: Caught signal number 11 SEGV: Segmentation Violation,
>>> probably memory access out of range
>>> > [25]PETSC ERROR: Try option -start_in_debugger or
>>> -on_error_attach_debugger
>>> > [25]PETSC ERROR: or see
>>> http://www.mcs.anl.gov/petsc/documentation/faq.html#valgrind
>>> > [25]PETSC ERROR: or try http://valgrind.org on GNU/linux and Apple
>>> Mac OS X to find memory corruption errors
>>> > [25]PETSC ERROR: likely location of problem given in stack below
>>> > [25]PETSC ERROR: ---------------------  Stack Frames
>>> ------------------------------------
>>> > [25]PETSC ERROR: Note: The EXACT line numbers in the stack are not
>>> available,
>>> > [25]PETSC ERROR:       INSTEAD the line number of the start of the
>>> function
>>> > [25]PETSC ERROR:       is given.
>>> > [25]PETSC ERROR: [25] SuperLU_DIST:pzgssvx line 236
>>> /home/cdfmat_marius/prog/petsc/master_build_gcc_local/petsc/src/mat/impls/aij/mpi/superlu_dist/superlu_dist.c
>>> > [25]PETSC ERROR: [25] MatMatSolve_SuperLU_DIST line 200
>>> /home/cdfmat_marius/prog/petsc/master_build_gcc_local/petsc/src/mat/impls/aij/mpi/superlu_dist/superlu_dist.c
>>> > [25]PETSC ERROR: [25] MatMatSolve line 3396
>>> /home/cdfmat_marius/prog/petsc/master_build_gcc_local/petsc/src/mat/interface/matrix.c
>>> > [25]PETSC ERROR: --------------------- Error Message
>>> --------------------------------------------------------------
>>> > [25]PETSC ERROR: Signal received
>>> > [25]PETSC ERROR: See
>>> http://www.mcs.anl.gov/petsc/documentation/faq.html for trouble
>>> shooting.
>>> > [25]PETSC ERROR: Petsc Development GIT revision: v3.10.2-723-g2a65a22
>>> GIT Date: 2018-11-25 07:45:29 +0100
>>> > [25]PETSC ERROR:
>>> /home/cdfmat_marius/prog/transomat_latest_test4/transomat on a  named a003
>>> by cdfmat_marius Thu Dec  6 10:12:23 2018
>>> > [25]PETSC ERROR: Configure options
>>> --prefix=/home/cdfmat_marius/prog/petsc/master_gcc_local_debug
>>> --with-64-bit-indices=0 --CC=mpicc --COPTFLAGS="-O0 -mtune=generic -g
>>> -std=c11 " --CXX=mpicxx --CXXOPTFLAGS="-O0 -mtune=generic -g -std=c++11 "
>>> --FC=mpif90 --FOPTFLAGS="-O0 -mtune=generic -g -fbacktrace -g "
>>> --with-mpi=1 --with-x=0
>>> --download-parmetis=/home/cdfmat_marius/prog/petsc/master_build_gcc_local/petsc/externalpackages/git.parmetis.tar.gz
>>> --download-metis=/home/cdfmat_marius/prog/petsc/master_build_gcc_local/petsc/externalpackages/git.metis.tar.gz
>>> --with-blaslapack-lib="
>>> -L/home/cdfmat_marius/intel/compilers_and_libraries_2019.1.144/linux/mkl/lib/intel64
>>> -Wl,--no-as-needed -lmkl_scalapack_lp64 -lmkl_intel_lp64 -lmkl_sequential
>>> -lmkl_core -lmkl_blacs_openmpi_lp64 -lpthread -lm -ldl "
>>> --with-scalapack-lib="
>>> -L/home/cdfmat_marius/intel/compilers_and_libraries_2019.1.144/linux/mkl/lib/intel64
>>> -Wl,--no-as-needed -lmkl_scalapack_lp64 -lmkl_intel_lp64 -lmkl_sequential
>>> -lmkl_core -lmkl_blacs_openmpi_lp64 -lpthread -lm -ldl "
>>> --download-pastix=no
>>> --download-superlu_dist=/home/cdfmat_marius/prog/petsc/master_build_gcc_local/petsc/externalpackages/git.superlu_dist.tar.gz
>>> --download-superlu_dist-commit=HEAD
>>> --download-hwloc=/home/cdfmat_marius/prog/petsc/master_build_gcc_local/petsc/externalpackages/hwloc-1.11.1.tar.gz
>>> --download-sowing=/home/cdfmat_marius/prog/petsc/master_build_gcc_local/petsc/externalpackages/git.sowing.tar.gz
>>> --with-mkl_pardiso-dir=/home/cdfmat_marius/intel/compilers_and_libraries_2019.1.144/linux/mkl
>>> --with-openmp=0 --with-pthread=0
>>> --download-elemental=/home/cdfmat_marius/prog/petsc/master_build_gcc_local/petsc/externalpackages/git.elemental.tar.gz
>>> --download-elemental-commit=HEAD
>>> --with-mkl_cpardiso-dir=/home/cdfmat_marius/intel/compilers_and_libraries_2019.1.144/linux/mkl
>>> --with-mkl_sparse_optimize-dir=/home/cdfmat_marius/intel/compilers_and_libraries_2019.1.144/linux/mkl
>>> --with-mkl_sparse_sp2m-dir=/home/cdfmat_marius/intel/compilers_and_libraries_2019.1.144/linux/mkl
>>> --with-cxx-dialect=c++11 --with-scalar-type=complex --with-debugging=1
>>> >
>>> > bestm
>>> > marius
>>>
>>
>>
>>
>> --
>> What most experimenters take for granted before they begin their
>> experiments is infinitely more interesting than any results to which their
>> experiments lead.
>> -- Norbert Wiener
>>
>> https://www.cse.buffalo.edu/~knepley/
>> <http://www.cse.buffalo.edu/~knepley/>
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-users/attachments/20181206/1cda9ba1/attachment.html>